Chip Gliedman, George Lawrie and John Rymer participated in a panel on packaged applications and Lean.
Rymer argued that packaged apps can never be Lean, since most are locked down, closed engines where the vendor controls the architecture, they’re expensive and difficult to upgrade, they use more functions than customers use, they provide a single general UI for all user personas, and each upgrade includes more crap that you don’t need. I tend to be on his side in this argument about some types of appls (as you might guess about someone who used to write code for a living), although I’m also a fan of buy over build because of that elusive promise of a lower TCO.
Gliedman argued the opposite side, pointing out that you just can’t build the level of functionality that a packaged application provides, and there can be data and integration issues once you abandon the wisdom of a single monolithic system that holds all your data and rules. I tend to agree with respect to functionality, such as process modeling: you really don’t want to build your own graphical process modeler, and the alternative is hacking your own process together using naked BPEL or some table-driven kludge. Custom coding also does not guarantee any sort of flexibility, since many changes may require significant development projects (if you write bad code, that is), rather than a package app that may be more configurable.
It’s never a 100% choice between packaged apps and custom development, however: you will always have some of each, and the key is finding the optimal mix. Lean packaged apps tend to be very fit-to-purpose, but that means that they become more like components or services than apps: I think that the key may be to look at composing apps from these Lean components rather than building Lean from scratch. Of course, that’s just service-oriented architecture, albeit with REST interfaces to SaaS services rather than SOAP interfaces to internal systems.
There are cases where Lean apps are completely sufficient for purpose, and we’re seeing a lot of that in the consumer Web 2.0 space. Consider Gmail as an alternative to an Exchange server (regardless of whether you use Outlook as a desktop client, which you can do with either): less functionality, but for most of us, it’s completely sufficient, and no footprint within an organization. SaaS, however, doesn’t not necessarily mean Lean. Also, there are a lot of Lean principles that can be applied to packaged application deployment, even if the app itself isn’t all that Lean: favoring modular applications; using open source; and using standards-based apps that fit into your architecture. Don’t build everything, just the things that provide your competitive differentiation where you can’t really do what you need in a packaged apps; for those things where you are doing the same all every other company, suck it up and consider a packaged app, even if it’s bulky.
Clearly, Gliedman is either insane or a secret plant from [insert large enterprise vendor name here], and Rymer is an incurable coder who probably has a ponytail tucked into his shirt collar. 🙂 Nonetheless, an entertaining discussion.